Research case
“Synthetic replacements for human survey data?” (Bisbee et al., Political Analysis 2024)
Political scientists asked ChatGPT to answer ANES feeling-thermometer questions as demographic personas and compared the results with real respondents.
What is reported
Averages looked close, but synthetic answers had far less variance, regressions often led to different conclusions, and results shifted with prompt wording and over three months.
Before you use it
A key counterweight to silicon-sample enthusiasm; read it before using simulated respondents for anything beyond pretesting.
Sources and review
Source checked · Not tested by this site
Checked the public abstract and bibliographic record. Results describe the authors' study, not a replication by this site.
License and reuse
No license has been recorded. Check the source’s license before reusing code, models or data.
A license note is not a license audit. Open source does not guarantee free hosting, model access or commercial reuse.